APBRmetrics

David Lee now has a 2009-10 (his last season in NYK) rating of 0.162 which ranks 77. I think this is a lot lower than it was before the revision.

Ah, indeed, going back to this article predicting 50 wins for the Warriors with Lee, it has Lee at 0.275 (a value which would have been top 10 this season). So it looks like it might have corrected a few of the players that people really had issues with (Lee, Humphries, Love a little bit, etc.).

I'm just glad the guy finally admitted something, although even in "defeat" he doesn't seem very gracious about it.

2010: 205 Dirk Nowitzki Dallas 4,3 3039 0.086 5.46
2009: 188 Dirk Nowitzki Dallas 4,0 3050 0.091 5.8

Do you honestly think that he improved the metric a tiny little bit? Nowitzki was in both years below league average according to WP48. Should someone really base a decision on WP48?

2010: 27 Troy Murphy Indiana 4,4 2344 0.212 10.36
2009: 17 Troy Murphy Indiana 4,0 2482 0.262 13.53

Seems like Troy Murphy was robbed by Nowitzki of TWO All-NBA 2nd awards.

Btw, by reducing the importance of defensive rebounding it also seems like they reduced the year to year consistency for the players.

mystic wrote:

DSMok1 wrote:I guess he's still not dealing with replacement level on rebounding?

No, he hasn't solved the real issue at all. The problem was never defensive rebounding alone. An important problem with his metric comes from the fact that his formula thinks a player can score points without using a possession. And while in reality the correlation coefficient between scoring margin and offensive rebounding over the last 10 years is around -0.2, Berri ends up with 0.2 as the correlation coefficient between offensive rebounding and WP48. When a player grabs an offensive rebound he used one possession less for his scoring according to Berri's formula used to calculate WP48. On a team basis offensive rebounding just keeps the possession, in Berri's world an offensive rebound is aquiring a possession.

The very basis of his approach is wrong. He is using formulas working on a team basis, but aren't useful for individuell players at all.

Agree with you on the team to player translations, but the correlation coefficient between scoring margin and offensive rebounding is just a product of multicollinearity. The greater the positive scoring margin (all else equal), the fewer the number of FG misses, and the fewer the offensive rebounding opportunities. Using a coeffcient whereby an offensive rebound generally cancels out a missed FG makes sense. But yeah, I think there are still a number of issues.

Replacement levels just don't make sense (see: the break even 2FG%). I don't lurk here to often, so please excuse me if this has been discussed a ton, but wouldn't it make sense for all of these linear weights formulas to apply a replacement value for ORB% as it relates to missed FGs.

The primary criticism of Hollinger's missed FG was that it was too generous with the DRB% assumption, thus forgiving the player of misses and allowing the break even 2FG% to be relatively low. Berri was the opposite.

Seeing as how the minimum team ORB rate in the 3 pt era is 20.9%, and how the minimum team DRB rate over the same period is roughly 61.5%, shouldn't the penalty for a FG miss somehow incorporate these in the penalty? Rebounding above these rates (at a team level) would allocate credit for the off. rebounder and that corresponds to the FG miss penalty.

Chicago76 wrote: Agree with you on the team to player translations, but the correlation coefficient between scoring margin and offensive rebounding is just a product of multicollinearity. The greater the positive scoring margin (all else equal), the fewer the number of FG misses, and the fewer the offensive rebounding opportunities. Using a coeffcient whereby an offensive rebound generally cancels out a missed FG makes sense. But yeah, I think there are still a number of issues.

Actually, that is not true. I used ORB%, thus the lesser amount of available rebounds is accounted for.

The fact is that in the last couple of seasons a lot of teams realized that having stretch bigs on the courts helps offensively a lot, but it also helps a lot, if those bigs are going back on defense quick rather than trying to get the offensive rebound. The fact is that the defending team has a better chance of getting the rebounding due to defensive positioning, the offense team has a disadvantage here, which leads to the fact that the chances are lower to get the rebound for the team on offense than for the team on defense. But having a big back on defense helps limiting the fastbreak opportunities for example. The Dallas Mavericks were an example for this, but also the Boston Celtics or San Antonio Spurs. The two best teams in ORB% were the Kings and the Timberwolves, two teams well below average on defense, while teams like the Celtics, Heat, Mavericks or Bucks were well above average on defense and well below average in ORB%.

Btw, an example of this:

Watch Nowitzki going back on defense right after he released the shot. That is a typical situation. You can find something similar with Garnett, when he is on top of key for example. Garnett didn't become a worse rebounder on offense, the amount of offensive rebounds became lower, because he went back on defense instead of going for the offensive rebound. And that also helped the Celtics defensively.

Chicago76 wrote: Seeing as how the minimum team ORB rate in the 3 pt era is 20.9%, and how the minimum team DRB rate over the same period is roughly 61.5%, shouldn't the penalty for a FG miss somehow incorporate these in the penalty? Rebounding above these rates (at a team level) would allocate credit for the off. rebounder and that corresponds to the FG miss penalty.

In ezPM a missed shot is worth PPP*ORB%, where ORB% is the league average.

For anyone interested in a long and sometimes heated discussion of break-even points, see the comments of my original ezPM article:

http://thecity2.com/2010/12/12/ezpm-yet ... valuation/

and my later post (somewhat cleverly) titled "Point Break...Even":

http://thecity2.com/2011/01/18/point-break-even/

mystic wrote:Actually, that is not true. I used ORB%, thus the lesser amount of available rebounds is accounted for.

That make sense. The way I read it, it sounded like you were working w/ raw numbers rather than rates.

mystic wrote:The fact is that in the last couple of seasons a lot of teams realized that having stretch bigs on the courts helps offensively a lot, but it also helps a lot, if those bigs are going back on defense quick rather than trying to get the offensive rebound. The fact is that the defending team has a better chance of getting the rebounding due to defensive positioning, the offense team has a disadvantage here, which leads to the fact that the chances are lower to get the rebound for the team on offense than for the team on defense.

I get this now. In other words: the transition defense cost of going for the ORB is greater than probability weighted benefit of retaining possession. I guess the only issue here is that if a player is sophisticated enough to know the trade-off, there are times when going for the ORB is still beneficial, so the neg coefficient shouldn't apply. That's just a hazard of linear weights: all situations where stats are accrued aren't equal. Kind of like the shot creator who is asked to put something up with the clock winding down and the lower eFG% that comes with that.

Thanks.

Dirk's rating on WP does suggest that there are some remaining metric issues.

Mystic's SPM had him 5th best last season. RAPM 2nd best. For some reasons in common and some different reasons too. RAPM credited his defense and probably his impact on the offensive effectiveness of others. SPM rated him high without addressing shot defense or impact on the offensive effectiveness of others (beyond assists).

Complete play by play based metrics like RAPM and EZPM have those advantages over boxscore metrics of all stripes even if these 2 still have their own issues.

I think a lot of Dirk's value comes from the mid-range shot, which taken in a vacuum, probably makes him seem less efficient. ezPM doesn't like Dirk either. That's one of the reasons I came up with the PSAMS metric, in which he's number one overall. I should probably figure out how to fold that into ezpm.

Chicago76 wrote: That make sense. The way I read it, it sounded like you were working w/ raw numbers rather than rates.

Ok, sorry, I should have clarified that before.

Chicago76 wrote:the transition defense cost of going for the ORB is greater than probability weighted benefit of retaining possession.

Correct, in most cases it is beneficial to just go back on defense. Fastbreak opportunities are one of the most efficient offensive opportunities, reducing them and forcing the opponents into halfcourt sets helps more than trying to get all rebounds on offense. Especially in the cases where the big is on the perimeter and due to his positioning has little to no chance to get the offensive rebound anyway.

Chicago76 wrote: I guess the only issue here is that if a player is sophisticated enough to know the trade-off, there are times when going for the ORB is still beneficial, so the neg coefficient shouldn't apply.

I agree. In my rating the offensive rebounds are incorporated like assists. An offensive rebounder gets a part of the points scored in each possession. By reducing the same amount of points from the non-offensive rebounder the overall points are not effected.

Chicago76 wrote: That's just a hazard of linear weights: all situations where stats are accrued aren't equal. Kind of like the shot creator who is asked to put something up with the clock winding down and the lower eFG% that comes with that.

Agree, especially with the limited informations the boxscore contains.

EvanZ wrote:ezPM doesn't like Dirk either.

I think there are two other reasons for this: First, the assisted FG value is lower (correct) and Nowitzki has somehow a pretty high AFG%. For whatever reason dumping the ball to Nowitzki into the midpost creates an assist, even though Nowitzki still has to work for his shot. Kidd gets probably 1 or 2 assists per game by just giving the ball to Nowitzki and then watching him making a couple of head and body fakes and taking that unorthodox one-legged fadeway. The other reason seems to be a lack of opportunity costs in regard to turnovers. Everytime a player takes a shot the team has a higher chance to score than letting the shot clock expire (well, in WP48 in a lot of cases it recommend to let the shot clock expire, because a shot clock violation is a team turnover, thus not credited to the player, while a forced shot to avoid a shot clock violation has a lower probability to go in than an open shot, that means the player would end up with a higher WP48). Last season for example the TS% was in average 54.1, that means in average 108.3 points were scored in 100 scoring attempts, that is better than the league average points per possession. That is not surprising at all, because when a team commits a turnover there will score exactly 0 points in that possession. Each shot attempt has a higher probability to create points than a turnover. And that is a huge reason why shot creation can't be ignored (like WP48 is basically doing it).

Hi, I'm a first time poster on the board, but have followed many of the discussions before. I just wanted to say how nice it is to see a thoughtful discussion about WP without insults being flung at anyone who questions the methodology or results. I have been working on my senior thesis for the last couple of months in which I refute some of Berri's findings, so I have spent a lot of time reading his papers and following his site. He is clearly a bright guy and I think he has done a lot to advance the field, but I can't stand his constant condescension towards anyone that questions WP. Sorry for not particularly adding to the discussion, but I just needed to vent about that. Thank you to everyone again, I really enjoy the board and hope to contribute more going forward.

Welcome, human.

Well, just took a look at WoW and found that nice piece in the comment section:

Alique Williams on January 15, 2012 at 3:40 am said:

@motherwell Dallas does use WP48 and not adjusted +- as much anymore.


Yeah, that explains perfectly them trading Humphries and giving Nowitzki a huge contract.



Evan, nice question and it is clear that Dre has no intention of answering it.

mystic wrote:


Evan, nice question and it is clear that Dre has no intention of answering it.

I know but I genuinely had to give him the opportunity. I thought maybe he was joking, but alas...

And that's not to say he's entirely wrong about Millsap being a very good, very underrated player. RAPM agrees (not that he's better than Dirk, but that he's a very good player). ezPM has Millsap rated ahead of Dirk, too, but I would never use that as the sole argument for claiming Millsap is a better player or that I would rather have Millsap over Dirk. It's truly a reality distortion zone over at that place. And it's annoying that some people (like "motherwell") don't realize there is a whole body of knowledge elsewhere on the interwebs that refutes WP quite easily.

EvanZ wrote: And that's not to say he's entirely wrong about Millsap being a very good, very underrated player. RAPM agrees (not that he's better than Dirk, but that he's a very good player). ezPM has Millsap rated ahead of Dirk, too, but I would never use that as the sole argument for claiming Millsap is a better player or that I would rather have Millsap over Dirk. It's truly a reality distortion zone over at that place.

I completely agree with everything. Millsap is a really good player. But solely basing your opinion about a player on a boxscore based stat, is insane. Not only is the boxscore biased towards offense, but even on offense it lacks a lot of informations.
And regarding Nowitzki it is really weird over there. Dre constantly avoids answers or gives answers like "Nowitzki was still 5th best player on the Mavericks during the finals" or "3rd best during the regular season". That Nowitzki in 2010 and 2009 was BELOW average in WP48 is something he doesn't say at all. And if Dallas is using WP48 as decision making tool, why did they sign a below average player to such a contract? Why did they trade away Humphries, who measured out better than Nowitzki in WP48?
They are using the trade for Odom as evidence, but Odom measured out pretty nicely by basically all metrics. My SPM had him in a similar range as WP48, 2yr APM had him at +5, RAPM at +3, all rather similar. Why not taking the signing of Vince Carter as indicator? Carter was constantly below average in WP48, last season at 0.062. Basing it on the conversion I can make with SPM to a wins per 48 minute value, I got Carter with 0.097. 2yr APM had Carter with +2.5, RAPM with +2. It sure as hell doesn't look like as if the Mavericks signed Carter, because Carter was doing great in WP48, but doing badly in +/- based metrics. In fact, it rather looks like as if the players, the Mavericks are going for, are really doing well in APM.
Even though they traded Brewer and Fernandez, but neither of those two were great WP48 players, but rather good in APM.

http://waynewinston.com/wordpress/?p=1213
http://waynewinston.com/wordpress/?p=1218

Winston explains very well, why the Mavericks probably traded for Fernandez. And I guess, the reasons for the trade of Brewer and Fernandez were the depth at the wing positions they had, the cap situation next season, both were the easiest to trade besides the top players, opening roster spots for depth at C.

How someone can look at the stuff the Mavericks are doing and conclude that they are using WP48 is beyond me. It looks like they are still paying good bucks for the stuff from Winston, they hired Beech as their stats guy, no idea, but that hardly sound like WP48 is a big key in their decision making.

EvanZ wrote: And it's annoying that some people (like "motherwell") don't realize there is a whole body of knowledge elsewhere on the interwebs that refutes WP quite easily.

True.

LA Blue Devil wrote:I have been working on my senior thesis for the last couple of months in which I refute some of Berri's findings

You are quite welcome here!

I can be seen frequently getting frustrated with Berri (like pretty much anyone else in the adjusted +/- game), but hopefully I haven't dropped too many ad hominems...